To generate confidence that recombinant CYPs may be used to predict HLM CLint data, the in vitro kinetics for commonly used CYP probes were determined previously in this laboratory in both pooled HLM and CYP coexpressed with NADPH-reductase in E. coli cells (McGinnity et al., 1999). The kinetic parameters (including CLint) of these recombinant enzymes were similar to their human liver counterparts for the enzyme substrate pairs that were directly comparable, and thus they would appear to be faithful surrogates. Indeed, Eddershaw and Dickins (1999) reported an excellent comparison between the rates of metabolism of several compounds determined from HLM and microsomes containing a mixture of the major recombinant CYPs. However, this "artificial HLM" approach gives little information as to the enzymology of metabolism.
To demonstrate the potential for predicting both the extent and route of oxidative metabolic clearance for NCEs by recombinant human CYPs, several marketed drugs were selected in which metabolism via CYP pathways was well established. Of the marketed drugs that are primarily cleared by human hepatic CYP-mediated metabolism, the vast majority were metabolized by one or more of the five isoforms, CYP1A2, -2C9, -2C19, -2D6, and -3A4 (Bertz and Granneman, 1997), and, for that reason, only these isoforms were employed in this initial study. There are limited available data on the relative levels of the five major isoforms in human hepatic microsomes (Table 2), and we have relied heavily on the seminal study by Shimada et al. (1994), which is widely cited for this purpose. The marker compounds tolbutamide, diazepam, metoprolol, ibuprofen, propranolol, dextromethorphan, omeprazole, diltiazem, testosterone, and verapamil were chosen so that metabolism by each of the five CYPs was adequately represented. The choice of DMSO as a solvent was based on its value for compounds with relatively low solubility (often encountered in early drug discovery programs) and its implementation as the solvent of choice in many liquid banks. Any inhibitory effects should not affect the comparison between recombinant CYPs and HLM.
There is a reasonable agreement between the CLint of the probe substrates determined in HLM to available literature values (Table 3), although the comparison is somewhat compromised due to the large spread of the literature data. There is agreement as to whether a compound demonstrates a low, intermediate, or high CLint. The limitations of such an interlaboratory comparison and the inherent variability of such an exercise are well established (Boobis et al., 1998). Literature CLint values have been obtained from a variety of sources, including Vmax/Km calculations, microsomes (prepared from individual as well as pooled livers), and hepatocytes [data converted to µl · min1mg1 assuming 2.67 × 106 cells/mg of microsomal protein (Carlile et al., 1999)]. Variability will also result from the fact that isoform levels and activities may vary significantly between the different metabolizing sources (Boobis et al., 1998). However, the agreement is excellent where a direct comparison between two laboratories determining HLM CLint for several compounds can be made (Obach, 1999).
Without exception, our data and the prevailing literature assign the same isoform to be the predominant CYP responsible for the metabolism of each marker compound. Diazepam at low micromolar concentrations was metabolized by CYP2C19, which agrees with Jung et al. (1997), Yasumori et al. (1993), and Andersson et al. (1994). Indeed, detailed HLM kinetics of diazepam metabolism (not shown) suggests the involvement of multiple CYPs (e.g., CYP2C9/18, -2B6, and -3A4), but the data indicate that the high affinity component of diazepam N-demethylation in vivo may be CYP2C19. Metoprolol and dextromethorphan are primarily CYP2D6 substrates (Otton et al., 1988; Dayer et al., 1989; Jacqz-Aigrain et al., 1993; Kerry et al., 1994; Von Moltke et al., 1998). Diltiazem, testosterone, and verapamil are predominantly metabolized by CYP3A4 (Waxman et al., 1988; Pichard et al., 1990; Kroemer et al., 1993; Sutton et al., 1997; Tracy et al., 1999).
In addition, for propranolol, omeprazole, tolbutamide, and ibuprofen, there was excellent concordance between our data and the literature on the relative contribution of several isoforms in the metabolism of the respective compounds. Tolbutamide is metabolized by both CYP2C9 (70%) and CYP2C19 (30%), which agrees with Inoue et al. (1997), Wester et al. (2000), Venkatakrishnan et al. (1998), and Lasker et al. (1998). Similarly, ibuprofen is metabolized by CYP2C9 (90%) and CYP2C19 (10%) (Leemann et al., 1993; Hamman et al., 1997) and omeprazole by CYP2C19 (68%) and CYP3A4 (32%) (Andersson et al., 1993; Karam et al., 1996; Yamazaki et al., 1997; Lasker et al., 1998). Indeed, when recombinant CYP2C19 and CYP3A4 were mixed at a ratio similar to that found in HLM, the metabolism of omeprazole resembled that of HLM (Yamazaki et al., 1997). The assignment of CYP2D6 (59%)-, CYP1A2 (26%)-, and CYP2C19 (15%)-dependent metabolism for propranolol agrees with several sources (Lennard et al., 1984; Otton et al., 1990; Yoshimoto et al., 1995), which implicate these three isoforms. In addition, the appropriate metabolites from each isoform were identified by mass spectrometry analysis.
A method based on the rate of enzyme activity (relative activity factor) of recombinant CYPs and HLM has proven useful in assigning the contribution of individual CYPs to several biotransformations (Rodrigues, 1999; Roy et al., 1999). Recently, it has been suggested that a complementary approach using the ratio of intrinsic clearance as a relative activity factor may be more predictive, where the kinetics for recombinant CYPs and HLM are equivalent (Nakajima et al., 1999). The correlation observed in this study between the sum of CLint from the different CYP isoforms and HLM CLint for the compounds tested confirms this concept. This study has additionally provided a more thorough evaluation of these recombinant proteins expressed in E. coli.
The summed CYP CLint correctly predicted a low HLM CLint (1mg1) for tolbutamide, diazepam, and metoprolol; an intermediate HLM CLint (8-65 µl · min1mg1) for ibuprofen, propranolol, dextromethorphan, diltiazem, and testosterone; and a high HLM CLint (>65 µl · min1mg1) for verapamil. However, the summed CYP CLint of omeprazole and propranolol did overpredict somewhat HLM CLint. One possible explanation for this is an increase in "futile" binding with increased protein concentration for some compounds. For propranolol there is 50% free at 0.4 mg · ml1 and 25% at 2 mg · ml1 (Obach, 1997), which results in a 2-fold decrease of propranolol CLint (Fig. 4). Typical assay conditions used 0.2 to 0.4 mg of protein/ml1 of CYPs (exact amount depended on the CYP expression level, because all incubations contain 100 pmol of CYP/ml1) and 1 mg · ml1 HLM. The HLM CLint of propranolol at 0.4 mg · ml1 was determined to be 22 ± 4 µl · min1mg1, which compares more favorably with the summed CYP CLint at the same protein level (55 ± 15 µl · min1mg1).
Generally, lower protein levels in the recombinant CYP assay may allow a more accurate reflection of unbound CLint and provide a greater dynamic CLint range when discriminating between large numbers of compounds. There is likely to be no significant differences between the extent of futile binding for HLM and recombinant CYPs at the same total protein concentration (Venkatakrishnan et al., 2000). Differential protein binding between in vitro matrices for predicting in vivo Clmet is currently under investigation.
In our experience, an accurate determination of a wide range of CLint is achieved at an incubation concentration for recombinant CYP of 100 pmol of CYP/ml1, which may be subsequently optimized. The molar ratio of NADPH-P450 reductase to recombinant CYP has been manipulated for the E. coli expression constructs to produce optimal reaction kinetics for probe substrates (McGinnity et al., 1999). For example, optimal CYP2C19-mediated diazepam N-demethylation can be achieved, in the absence of cytochrome b5, by increasing the molar ratio of NADPH-P450 reductase:CYP2C19 to approximately 20:1 (McGinnity et al., 1999). Indeed, to optimize CYP expression systems, further elucidation of the role and importance of ancillary electron transporters such as b5 in the metabolism of xenobiotics is required (Yamazaki et al., 1999).
A correlation between HLM and CYP CLint allows compounds to be ranked with respect to metabolic stability, should expedite knowledge of the pharmacophore of individual CYP isoforms, and may facilitate more rational compound synthesis to achieve greater metabolic stability. Furthermore, an underprediction of HLM CLint from the five major human hepatic isoforms should prompt an investigation into possible metabolism by the more minor human hepatic CYPs, i.e., CYP2A6, -2B6, -2C8, or -2E1 (Houston, 1994).
This automated assay is being used early in drug discovery at AstraZeneca R&D Charnwood, a strategy distinct from the comprehensive isoform profiling of a drug later in the development process by other groups (Machinist et al., 1998; Fischer et al., 1999; Nakajima et al., 1999; Roy et al., 1999). The early identification of the major CYP isoforms involved in the metabolism of a drug candidate is useful for several purposes, including understanding ligand-enzyme structure-activity relationships, expanding the database for substrates of the polymorphic isoforms, assessing the potential intersubject variability, and predicting the drug-drug interactions and, ultimately, the direction of clinical trials.
These data indicate that recombinant CYPs may be used to predict HLM CLint. Furthermore, it may prove feasible to scale human CLint data to the fractional metabolic clearance encountered clinically (Iwatsubo et al., 1997; Obach, 1999). Therefore, although very much in its infancy, data in this report demonstrate that E. coli-expressed CYPs may be useful as an early approach for the prediction of the enzymology of human CYP metabolism. Further efforts to examine the differential nonspecific binding between the separate in vitro models and the effects on CLint are underway.
Received May 24, 2000; accepted August 7, 2000.
1 Current address: Department of Molecular and Cellular Pathology, Ninewells Hospital and Medical School, Dundee DD1 9SY, UK.
3 Copies of the program are available from the corresponding author upon request.
Send reprint requests to: Dr. Rob Riley, Department of Physical & Metabolic Science, AstraZeneca R&D Charnwood, Loughborough, Leicestershire LE11 5RH, UK. E-mail: [email protected]
Abbreviations used are: CYP, cytochrome P450; HLM, human liver microsomes; CLint, intrinsic clearance; NCE, new chemical entity; RSP, robotic sample processor.